1 / 14

kernel development and the embedded world

Andrew Morton <akpm@linux-foundation.org> <akpm@google.com> CELF Embedded Linux Conference April 2008. kernel.org development and the embedded world. Overview. Assumed audience is developers of embedded products rather than developers of Linux itself

naava
Download Presentation

kernel development and the embedded world

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Andrew Morton <akpm@linux-foundation.org> <akpm@google.com> CELF Embedded Linux Conference April 2008 kernel.org development and the embedded world

  2. Overview • Assumed audience is developers of embedded products rather than developers of Linux itself • Won't cover day-to-day interaction with the kernel.org team. • Will cover economics of embedded, how this impacts their kernel.org participation • Patch-hoarding, and why not to do it • Why merge upstream? • Which kernel version to use? • Why do kernel.org developers give free support? • The merge decision making processes

  3. The economics of Linux-for-embedded • The product's kernel version is usually not upgraded across product lifetime • In desktop and server kernel upgrades are expected • In embedded the software reaches the end-user directly from the manufacturer • In desktop and server, it is usually via intermediaries • Exceptional case: hardware manufacturers whose customers develop embedded products • There seems to be generally less cash handy in the embedded segment • Net effect: embedded development is underrepresented in kernel.org work, relative to its economic importance

  4. Effects of embedded's under-representation • There is a risk that it will tilt development too far toward server/desktop • However the kernel.org team care about Linux-on-embedded and want it to work well • So we do consider each change's effects on small systems • However this is mainly a minimize-the-damage effort • New server/desktop features should not degrade embedded • But only a few developers are actively working to improve Linux for embedded • This is unfortunate – just a single full-time “embedded maintainer” could have a large effect.

  5. How you can help kernel.org development • Not all contributions are code patches! • We (and the kernel) can benefit from your experience • Tell us your problems • Which features are missing • Help us prioritize different proposals/features • Help us to direct those resources which we do have • Read the websites, read the mailing lists and don't just lurk • Review patches • Suggesting improvements is a good way to get work done for free • The squeaky gate gets the oil

  6. Patch hoarding • Various embedded groups seem to like to maintain out-of-tree patchsets • These patchsets linger on for ever and never seem to get merged • Due to lack of resources (I hope)‏ • To retain a little competitive advantage (I hope not)‏ • It is inefficient and ineffective to have all these little patchpiles dotted around the place • Let's get them merged up

  7. Why merge upstream? The pros • The review, integration and release process will improve the code quality • Once merged, your maintenance cost falls to near-zero • Others will improve it • Others will no longer break it (as much)‏ • Others will enhance it • It gets more testing • Others will fix bugs in it • Eliminates possibility that a competing implementation will be merged • You avoid the risk of having to migrate to a new implementation, or of maintaining yours for ever

  8. Why merge upstream? The cons • It requires some up-front work • But less long-term work (I think)‏ • You may be asked to change things • This hurts if you've already deployed it • This is a reason to merge before deploying! • The feature becomes more easily available to your competitors • Tough luck, live with it.

  9. Which kernel version to use? • Many embedded developers use old kernels • And that's OK – they were fine kernels • Much of the development in recent years has been addressing large-system shortcomings • However, freezing on an old kernel is a short-term strategy. • When the time comes to upgrade it can be painful • And you will want to upgrade, to pick up new features • Using an old kernel rules out support from the kernel.org team • A better strategy might be to develop product on current kernel.org release, freeze it at end of product development • Intermediate kernel providers should be able to help • kernel.org developers might be able to help

  10. Which kernel version to use? (cont'd)‏ • A single-product company can feasibly freeze on one kernel version for ever • This is probably less viable for multi-product companies

  11. Free support!! • Why do kernel.org developers give free support? • Well, we don't. • We will work with you to resolve kernel problems and shortcomings • It is a mutually beneficial transaction • If fixing your problem won't improve the mainline kernel we won't help you (much)‏ • If your problem is in 2.6.ancient then we won't be very interested • First question: “does it happen in current mainline”? • This is a strong reason for using recent kernels in product development

  12. The merge decision • The “dictator model”: maintainers make arbitrary decisions • We don't do this • The “rule of law” model: maintainers decide whether a submission meets the written rules • Maintainer is a “judge” and not a “dictator” • We do do this • Unfortunately there are no written rules ;)‏ • Should meet coding guidelines (these are written)‏ • Is useful to some kernel users • Long-term maintainable (traded off against usefulness)‏ • Non-injurious to other parties <- most important? • If the feature is large: does it have a maintenance team?

  13. The merge decision (cont'd)‏ • When in doubt: ask me! • I can help • That's what I'm here for • I'm really nice! • And I have great legs • But I'll only help you if there's something in it for me

  14. Working with the kernel.org team • Identify one or two specialists in your organization as the kernel.org interface • Don't expect every developer to work with the public developers • It just isn't efficient • Or effective • When in doubt, ask me.

More Related